We introduce NExT-QA, a rigorously designed video question answering (VideoQA) benchmark to advance video understanding from describing to explaining the temporal actions. Based on the dataset, we set up multi-choice and open-ended QA tasks targeting causal action reasoning, temporal action reasoning, and common scene comprehension. Through extensive analysis of baselines and established VideoQA techniques, we find that top-performing methods excel at shallow scene descriptions but are weak in causal and temporal action reasoning. Furthermore, the models that are effective on multi-choice QA, when adapted to open-ended QA, still struggle in generalizing the answers. This raises doubt on the ability of these models to reason and highlights possibilities for improvement. With detailed results for different question types and heuristic observations for future works, we hope NExT-QA will guide the next generation of VQA research to go beyond superficial scene description towards a deeper understanding of videos. (The dataset and related resources are available at https://github.com/doc-doc/NExT-QA.git)
Temporal convolutional networks (TCNs) are a commonly used architecture for temporal video segmentation. TCNs however, tend to suffer from over-segmentation errors and require additional refinement modules to ensure smoothness and temporal coherency. In this work, we propose a novel temporal encoder-decoder to tackle the problem of sequence fragmentation. In particular, the decoder follows a coarse-to-fine structure with an implicit ensemble of multiple temporal resolutions. The ensembling produces smoother segmentations that are more accurate and better-calibrated, bypassing the need for additional refinement modules. In addition, we enhance our training with a multi-resolution feature-augmentation strategy to promote robustness to varying temporal resolutions. Finally, to support our architecture and encourage further sequence coherency, we propose an action loss that penalizes misclassifications at the video level. Experiments show that our stand-alone architecture, together with our novel feature-augmentation strategy and new loss, outperforms the state-of-the-art on three temporal video segmentation benchmarks.
Segmenting objects of interest in an image is an essential building block of applications such as photo-editing and image analysis. Under interactive settings, one should achieve good segmentations while minimizing user input. Current deep learning-based interactive segmentation approaches use early fusion and incorporate user cues at the image input layer. Since segmentation CNNs have many layers, early fusion may weaken the influence of user interactions on the final prediction results. As such, we propose a new multi-stage guidance framework for interactive segmentation. By incorporating user cues at different stages of the network, we allow user interactions to impact the final segmentation output in a more direct way. Our proposed framework has a negligible increase in parameter count compared to early-fusion frameworks. We perform extensive experimentation on the standard interactive instance segmentation and one-click segmentation benchmarks and report state-of-the-art performance.
In current interactive instance segmentation works, the user is granted a free hand when providing clicks to segment an object; clicks are allowed on background pixels and other object instances far from the target object. This form of interaction is highly inconsistent with the end goal of efficiently isolating objects of interest. In our work, we propose a clicking scheme wherein user interactions are restricted to the proximity of the object. In addition, we propose a novel transformation of the user-provided clicks to generate a weak localization prior on the object which is consistent with image structures such as edges, textures etc. We demonstrate the effectiveness of our proposed clicking scheme and localization strategy through detailed experimentation in which we raise state-of-the-art on several standard interactive segmentation benchmarks.
In this paper, we show that ImageNet-Pretrained standard deep CNN models can be used as strong baseline networks for audio classification. Even though there is a significant difference between audio Spectrogram and standard ImageNet image samples, transfer learning assumptions still hold firmly. To understand what enables the ImageNet pretrained models to learn useful audio representations, we systematically study how much of pretrained weights is useful for learning spectrograms. We show (1) that for a given standard model using pretrained weights is better than using randomly initialized weights (2) qualitative results of what the CNNs learn from the spectrograms by visualizing the gradients. Besides, we show that even though we use the pretrained model weights for initialization, there is variance in performance in various output runs of the same model. This variance in performance is due to the random initialization of linear classification layer and random mini-batch orderings in multiple runs. This brings significant diversity to build stronger ensemble models with an overall improvement in accuracy. An ensemble of ImageNet pretrained DenseNet achieves 92.89% validation accuracy on the ESC-50 dataset and 87.42% validation accuracy on the UrbanSound8K dataset which is the current state-of-the-art on both of these datasets.
Future prediction requires reasoning from current and past observations and raises several fundamental questions. How much past information is necessary? What is a reasonable temporal scale to process the past? How much semantic abstraction is required? We address all of these questions with a flexible multi-granular temporal aggregation framework. We show that it is possible to achieve state-of-the-art results in both next action and dense anticipation using simple techniques such as max pooling and attention. To demonstrate the anticipation capabilities of our model, we conduct experiments on the Breakfast Actions, 50Salads and EPIC-Kitchens datasets where we achieve state-of-the-art or comparable results. We also show that our model can be used for temporal video segmentation and action recognition with minimal modifications.
Wavelets are well known for data compression, yet have rarely been applied to the compression of neural networks. In this paper, we show how the fast wavelet transform can be applied to compress linear layers in neural networks. Linear layers still occupy a significant portion of the parameters in recurrent neural networks (RNNs). Through our method, we can learn both the wavelet bases as well as corresponding coefficients to efficiently represent the linear layers of RNNs. Our wavelet compressed RNNs have significantly fewer parameters yet still perform competitively with state-of-the-art on both synthetic and real-world RNN benchmarks. Wavelet optimization adds basis flexibility, without large numbers of extra weights